Rejecting unknown values (e.g. sequencing platform) creates unnecessary extra work

[fiendish]

Data value set membership enforcement is probably a mistake. Adding a sequencing experiment that was run on a new platform shouldn't require an upgrade of the dataservice. That's a lot of extra work for just some new data.

It seems like the implementation of "known values should be easily enumerated so that we can all be consistent" somehow turned into "we should restrict input to only things that we've seen before", but that does not seem like an appropriate interpretation to me.